Since the accurate description of vowels by type in the Korean language is challenging and it is difficult to ensure objectivity, vowels have not been studied to the same degree as have consonants in acoustic phonetic research on Korean. In particular, children’s vowel expressions have not been widely analyzed due to the broad frequencies, impact of nasalization, and various voice qualities involved. However, related disabilities can be identified by determining values for the acoustic variables of vowels and objectively evaluated by a formant analysis of vowel articulation errors. Furthermore, the formant values for the positions of vowel articulation can be used to determine the impact of articulation therapy. This research presents two setting parameters for the formant analysis of children’s vowels and creates an automated script allowing optimal formant analysis.
Using Burg’s method in Praat, formants can be extracted by setting the maximum formant value and the number of formants. In this paper, the optimal formant setting was determined by identifying the two conditions, F1 and F2, with minimum standard deviations.
When applying the optimal formant setting determined by implementing the script presented in this study, the results of normality tests were not significant among all vowels except /u/ for the maximum formant value, and among the vowels /a/, /e/, /i/, /o/, /?/, and /?/ for the number of formants. All vowels except /o/ regarding the maximum formant value and all vowels except /i/ regarding the number of formants exhibited kurtosis in the sound and a platykurtic distribution. This indicates that when analyzing the formants of children’s vowel sounds, the unilateral application of a parameter setting (the maximum formant value and the number of formants) to all sound data is problematic.
The smaller standard deviations of F1 and F2 in the formant analysis indicate that the formants have been extracted more appropriately. According to a Kruskal-Wallis test, the differences between vowel groups were statistically significant for both F1_SD and F2_SD. In Tukey’s HSD test, the scale of deviation for F1_SD increased in the order of /?/, /u/, /i/, /o/, /?/, /e/, and /a/. Based on the large difference between [?, u, i, o] and [?, e, a], the vowels could be segmented into two groups. In Fe_SD, the scale of deviation grew in the order of /?/, /o/, /u/, /e/, /a/, /?/, and /i/. Based on their scale of deviation, they were grouped into [?, o, u, e] and [a, ?, i]. This suggests that a larger mouth opening results in more difficult articulation and unstable energy, and consequently a greater deviation in F1. To the contrary, with a smaller mouth opening articulation is achieved in a more stable state with the upper and lower teeth remaining closer. As a result, a smaller deviation in F1 is generated. In other words, the larger the mouth opening, the greater the deviation in F1. As for F2, the deviation was greatest in the group [a, ?, i]. As established in previous research, the deviations of /a/ and /?/, both of which have greater frontness among back vowels in Korean, were greater. This finding leads to the assumption that vowels with greater frontness have a greater deviation in F2. Furthermore, the vowels with greater deviations in F2 were those with a larger or smaller mouth opening compared to other vowels. This allows the assumption that deviation in F2 increases among high or low vowels with the tendency of frontness.
Finally, the performance of the optimal formant setting script was evaluated in order to determine whether it properly extracts formants for children’s vowels. To this end, the vowels of 15 6-year-old children were collected and the script was applied to the data. The results were judged by seven evaluators using a 5-point Likert scale. The congruence between the formant line and spectrogram was 99.9% with F1 and 92.971% with F2. In other words, the script performed very well in formant extraction.
For this research, an automated formant setting script designed for optimal formant analysis of children’s vowels was created separately for researchers and for users. In the case of the script for researchers, it was designed to automatically conduct optimal formant analysis for a large volume of child sound data and present the results in a text file. The script for users was intended to perform an immediate formant analysis of data obtained in a clinic. This will allow the clinical use of formant measures for analyzing articulation errors of vowels or as an indicator for the effectiveness of therapy.
While the script presented in this research was found to actualize an optimal setting and show a satisfying performance in the formant analysis of children''s vowels, both the sample size and the age group for the study were limited. There is a need for subsequent research evaluating the performance of the script by expanding the sample size and the ages of the subject group. Further research is required on diphthongs in the Korean language. Last, the formant setting script for children may be modified in order to apply it to formant analysis of the vowels of a broader range of ages.